A Flexible Solution-Processed Memristor

Home , Printed electronics

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE ELECTRON DEVICE LETTERS 1 A Flexible Solution-Processed Memristor Nadine Gergel-Hackett, Member, IEEE, Behrang Hamadani, Barbara Dunlap, John Suehle, Senior Member, IEEE, Curt Richter, Senior Member, IEEE, Christina Hacker, and David Gundlach, Member, IEEE

Abstract—A rewriteable low-power operation nonvolatile physically flexible memristor device is demonstrated. The active compo- nent of the device is inexpensively fabricated at room temperature by spinning a TiO2 sol gel on a commercially available polymer sheet. The device exhibits memory behavior consistent with a memristor, demonstrates an on/off ratio greater than 10 000 : 1, is nonvolatile for over 1.2 × 106 s, requires less than 10 V, and is still operational after being physically flexed more than 4000 times. Index Terms—Flexible electronics, flexible memory, memristor, sol gel, titanium dioxide.

I. INTRODUCTION E HAVE fabricated a physically flexible solution- W processed device that exhibits electrical behavior con- Fig. 1. Flexible polymer sheet patterned with four rewriteable nonvolatile sistent with that of a memristor, a memory device recently flexible TiO2 sol gel memory devices with cross-bar aluminum contacts. The experimentally demonstrated and proposed to be the miss- inset is a side view cartoon of the flexible TiO2 device structure. ing fourth basic circuit element [1], [2]. Although electrical make our flexible memristor device a prime candidate for use switching behavior has been observed from organic mono- in inexpensive flexible lightweight portable electronics, such as layers and metal oxides from as early 1968 [3]–[11], the disposable sensors [13]–[17]. unique electrical characteristics associated with the memristor have the potential to revolutionize computing. For example, the functionality of a memristor has been compared to that II. EXPERIMENT of a neurological synapse; thus, memristors may eventually Rather than using expensive methods and equipment for the enable electronic computation functionally similar that which deposition of the active TiO2 layer of our device, we per- occurs in the human brain [12]. We have fabricated devices formed a room-temperature deposition through a spin-on sol gel that demonstrate these desirable memristor characteristics as process that required no annealing [18]. This procedure consists well as: physical flexibility, fabrication by using inexpensive of spinning a titanium isopropoxide solution on the flexible room-temperature solution processing, operation voltages of plastic substrate (spun on at a rate of approximately 33 r/s for less than 10 V (relatively low power for flexible electronics), 60 s), and then leaving the precursor in air for at least 1 h to on/off ratios greater than 10 000 : 1, memory potential that is hydrolyze and form a 60-nm-thick amorphous TiO film [18]. × 6 2 nonvolatile for over 1.2 10 s, and the ability to be operated To electrically contact the active area, a simple two-terminal after being physically flexed 4000 times. These characteristics crossbar is formed by depositing the bottom contact (80 nm Al) on the substrate (approximately 2.5 cm × 2.5 cm square of HP Manuscript received March 18, 2009; revised April 6, 2009. This work color laserjet transparency C2934A) prior to spinning the pre- was supported in part by the NIST Office of Microelectronics Programs, by the NIST Competence on Organic Electronics, and by the DARPA MoleApps cursor, and depositing the top contact (80 nm Al) after the pre- Program. The research was performed while the first two authors held National cursor has hydrolyzed. The devices presented in this letter were Research Council Research Associate Awards and the third author was a fabricated by depositing the top and bottom metal contacts via Society of Physics Students Summer Intern. Certain equipment, instruments, thermal evaporation through a shadow mask; however, there is or materials are identified in this paper in order to adequately specify the experimental details. Such identification neither implies recommendation by the potential to deposit contacts from solution to enable roll-to- the National Institute of Standards and Technology nor does it imply the roll or inkjet processing [19]. The final Al/TiO2/Al stack has materials are necessarily the best available for the purpose. The review of this an area of approximately 2 mm × 2mmandisshowninFig.1. letter was arranged by Editor C.-P. Chang. N. Gergel-Hackett, B. Hamadani, J. Suehle, C. Richter, C. Hacker, and D. Gundlach are with the National Institute of Standards and III. RESULTS AND DISCUSSION Technology, Gaithersburg, MD 20899 USA (e-mail: [email protected]; [email protected]; [email protected]; [email protected]; The flexible TiO2-based devices exhibit electrical switching [email protected]; [email protected]). B. Dunlap was with the National Institute of Standards and Technology, with memory characteristics that match the electrical behavior Gaithersburg, MD 20899 USA. She is now with Alleghany College, Meadville, reported for memristors [1], [2]. A representative example of PA 16335 USA (e-mail: [email protected]). the basic switching behavior is shown in Fig. 2. On the first Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. voltage sweep (1), the device starts in the low-conductivity state Digital Object Identifier 10.1109/LED.2009.2021418 until the bias is increased to approximately +3 V, when the

Authorized licensed use limited to: NIST Researchers. Downloaded on June 3, 2009 at 11:25 from IEEE Xplore. Restrictions apply. This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE ELECTRON DEVICE LETTERS

Fig. 2. Current versus voltage for one switching cycle of a device. The inset shows sweeps of the same bias polarity applied consecutively (i.e., consecutive “writes” with no “erasing”) to a previously unbiased device, demonstrating a decrease in device resistance with each sweep as the device transitions from a low to high state in multiple steps. The electrical results shown in the inset Fig. 3. Median current at 0.25 V from several devices initially set in the are from a different device than those shown in the body of the figure. For “high” state and “read” (by applying 0.25 V) at various time increments, and measurements, the current compliance was 50 mA, and the voltage compliance the median current from several devices initially set in the “low” state and then was 10 V. “read” at various time increments. Moreover, shown on the graph is the standard deviation of the current values “read” at all elapsed times for a representative current abruptly transitions to a high conductivity state device initially set in the “high” state and for a representative device initially (“write”). This state remains until a negative bias of approx- set in the “low” state. imately −2 V is applied (2), resulting in a transition back to a low-conductivity state (“erase”). The initial “write” of a Although our TiO2 film was deposited by using a nontra- device can be achieved by applying either an adequate positive ditional sol gel method, the unbiased film was determined via or adequate negative bias through a voltage sweep or pulse; X-ray photoelectron spectroscopy to have the same chemical however, once this initial switch is performed, the opposite composition as that of reported memristors [22]. Additionally, polarity is required to “erase.” These devices also showed although there has been some question as to the origin of relatively high yields. Of the 50 devices that were fabricated switching in devices with aluminum contacts [23], the switch- over seven different experimental runs, 42 devices exhibited ing behavior from our devices cannot be attributed to the electrical behavior indicative of electrical switching (overall aluminum (or aluminum oxide), since it was also observed from yield of 84%) with the 16% of devices that did not show devices with noble metal (gold, silver, and platinum) contacts. switching behavior as electrical “shorts” (R<50 Ω).The Furthermore, the devices still exhibited switching behavior 50 switching devices exhibited a median “write” bias mag- when placed under vacuum overnight and then electrically nitude of |2.8|±1.3 V, a median “erase” bias magnitude of characterized in an argon environment, which indicates that |4.0|±1.4 V, a median high state resistance of 1.1 × 103 Ω, the behavior is not ambient dependent. The switching was and a median low state resistance of 2.7 × 107 Ω. We performed also observed when a glass substrate was used instead of a multiple switching cycles on devices and observed switching polymer sheet, indicating that while using the polymer sheet between high and low states with resistance values similar to demonstrates the potential of the devices to be used as flexible those of the original cycles. substrates, the switching mechanism is not intrinsic to it. These electrical results are consistent with the theorized The memory of the devices remained nonvolatile for more memristor mechanism, the drift of oxygen vacancies in the than 1.2 × 106 s (14 days). To characterize the volatility of the TiO2 under bias [1]–[3], [20]. Although it is possible that this flexible memristors, several devices were first each set “high” vacancy drift occurs through a single filament rather than a or “low” by applying voltage sweeps of the appropriate polar- frontlike system, filament switching generally results in binary ity until the device resistance was consistent with previously electrical characteristics [4]. While our devices can be operated observed values for the respective state. These states were then as binary switches (“hard switching”), they can also be operated “read” (by applying 0.25 V) on day 0, and periodically over in the “soft switching” regime where consecutive bias sweeps time. Fig. 3 shows the median current “read” over time from the of the same polarity result in a step-down in resistance with devices initially set “high,” as well as the median “read” over each sweep. For example, the inset of Fig. 2 demonstrates “soft time from the devices initially set in “low.” The error bars to the switching” with voltage sweeps of the same polarity consec- right of the graph indicate the standard deviation of the current utively applied to a previously unbiased device (i.e., multiple over the same time scale for a representative device initially “writes” applied in a row without “erasing” in-between). As set in the “high” state and for a representative device initially these consecutive sweeps are applied, the device transitions set in the “low” state. As these data show, the average device from a low to high state in multiple steps, which is consistent demonstrates no noticeable change in the current magnitude of with the simulated theoretical electrical behavior of a memristor the set state even after more than 1.2 × 106 s (14 days). in the “soft switching” regime [1]. The complete details of the To demonstrate mechanical robustness, the devices were switching mechanism of the memristor are explored in great electrically characterized after being physically flexed with detail elsewhere [1], [2], [21]. an automated flexing apparatus. The flexing apparatus gripped

GERGEL-HACKETT et al.: FLEXIBLE SOLUTION-PROCESSED MEMRISTOR 3

REFERENCES [1] D. B. Strukov, G. S. Snider, and D. R. Stewart, “The missing memristor found,” Nature, vol. 453, no. 7191, pp. 80–83, May 2008. [2] J. J. Yang, M. D. Pickett, X. Li, D. A. A. Ohlberg, D. R. Stewart, and R. S. Williams, “Memristive switching mechanism for metal/oxide/ metal nanodevices,” Nat. Nanotechnol., vol. 3, no. 7, pp. 429–433, Jul. 2008. [3] F. Argall, “Switching phenomena in titanium oxide thin films,” Solid State Electron., vol. 11, no. 5, pp. 535–541, May 1968. [4] A. A. Ansari and A. Qadeer, “Memory switching in thermally grown titanium oxide films,” J.Phys.D,Appl.Phys., vol. 18, no. 5, pp. 911– 917, May 1985. [5] A. Beck, J. G. Bednortz, C. Gerber, C. Rossel, and D. Widmer, “Repro- ducible switching effect in thin oxide films for memory applications,” Appl. Phys. Lett., vol. 77, no. 1, pp. 139–141, Jul. 2000. [6] I. Park, K. Kim, S. Lee, and J. Ahn, “Resistive switching characteristics for nonvolatile memory operation of binary metal oxides,” Jpn. J. Appl. Fig. 4. Representative “read” current versus voltage for the devices after being Phys., vol. 46, no. 4B, pp. 2172–2174, 2007. set into a high current state. The current states were “read” after 0 flexes, [7] P. J. Kuekes, D. R. Stewart, and R. Stanley Williams, “The cross- 100 flexes, 2000 flexes, 3000 flexes, and 4000 flexes (total), and repeated after bar latch: Logic value storage, restoration, and inversion in crossbar the devices were set low. The ratio of the high state current to low state current, circuits,” J. Appl. Phys., vol. 97, no. 3, pp. 034 301.1–034 301.3, for high and low states flexed an equal number of cycles after being set, remains Feb. 2005. at least 10 000 : 1. The inset shows a cycle of switching for a device that has [8] M. A. Reed, L. Chen, A. M. Rawlett, D. W. Price, and J. M. Tour, “Mole- been flexed more than 4000 times, demonstrating that the devices are still cular random access memory cell,” Appl. Phys. Lett., vol. 78, no. 23, operational after switching. pp. 3735–3737, Jun. 2001. [9] N. Majumdar, N. Gergel, G. Pattanaik, G. Zangari, Y. Yao, J. M. Tour, the edges of the sample and flexed the device from flat to a J. C. Bean, and L. R. Harriott, “The electrical behavior of nitro half-ellipse with a semiminor axis of 2.5 mm and a semimajor oligo(phenylene ethynylene)’s in pure and mixed monolayers,” J. Elec- axis of 8.5 mm. For the flexing measurements, several devices tron. Mater., vol. 35, no. 1, pp. 140–146, Jan. 2006. [10] J. A. Green, J. W. Choi, A. Boukai, Y. Bunimovich, E. Johnston-Halperin, were “written,” flexed 100 times, and then the current state was E. DeIonno, Y. Luo, B. A. Sheriff, K. Xu, Y. S. Shin, H. Tseng, read by applying a voltage of 0.25 V. The current state was then J. F. Stoddart, and J. R. Heath, “A 160-kilobit molecular electronic read again after 1900 more flexes (2000 total), and then again memory patterned at 1011 bits per square centimeter,” Nature, vol. 445, no. 7126, pp. 414–417, Jan. 2007. after every 1000 flexes until the devices were each flexed for a [11] C. A. Richter, D. R. Stewart, D. A. A. Ohlberg, and R. Stanly Williams, total of 4000 times. At this point, an adequate negative bias was “Electrical characterization of Al/AlOx/molecule/Ti/Al devices,” applied to each device to switch it into the low state (erased), Appl. Phys. A—Mater. Sci. Process., vol. 80, no. 6, pp. 1355–1362, Mar. 2005. and the flex/read cycles described above were repeated. Fig. 4 [12] R. Williams, “How we found the missing memristor,” IEEE Spectr., shows representative read current versus voltage for the devices vol. 45, no. 12, pp. 28–35, Dec. 2008. after flexing. As evident in this graph, both the high and low [13] M. G. Kane, J. Campi, M. S. Hammond, F. P. Cuomo, B. Greening, C. D. Sheraw, J. A. Nichols, D. J. Gundlach, J. R. Huang, C. C. Kuo, states decreased by about half of an order of magnitude with L. Jia, H. Klauk, and T. N. Jackson, “Analog and digital circuits using or- each 1000 flexes. Yet, the ratio of the high state current to low ganic thin-film transistors on polyester substrates,” IEEE Electron Device state current, for high and low states flexed an equal number Lett., vol. 21, no. 11, pp. 534–536, Nov. 2000. [14] D. Gomota-Chair, “Organic and printed electronics,” iNEMI Technology of cycles after being set, remains at least 10 000 : 1. This Roadmaps, Jan. 2007. demonstrates that even after flexing 4000 times the devices [15] S. Moller, C. Perlov, W. Jackson, C. Taussig, and S. R. Forrest, “A poly- retain an adequate state ratio for memory applications. The mer/semiconductor write-once read-many-times memory,” Lett. Nature, vol. 426, no. 6963, pp. 166–169, Nov. 2003. inset of Fig. 4 is a cycle of switching for a device that has been [16] L. Li, Q. Ling, S. Lim, Y. Tan, C. Zhu, D. S. H. Chan, E. Kang, and flexed more than 4000 times, demonstrating that the devices K. Neoh, “A flexible polymer memory device,” Organic Electron.,vol.8, are still operational after significant flexing. no. 4, pp. 401–406, Aug. 2007. [17] K. Asadi, D. M. De Leeuw, B. De Boer, and P. W. M. Blom, “Organic non-volatile memories from ferroelectric phase-separated blends,” Nat. IV. CONCLUSION Mater., vol. 7, no. 7, pp. 547–550, Jul. 2008. [18] J. Y. Kim, S. H. Kim, H. Lee, K. Lee, W. Ma, X. Gong, and A. J. Heeger, We have described the fabrication and electrical behavior “New architecture for high-efficiency polymer photovoltaic cells using of a simple inexpensive solution-processed low-power device solution-based titanium oxide as an optical spare,” Adv. Mater., vol. 18, no. 5, pp. 572–576, Mar. 2006. for flexible nonvolatile memory. Prototypical devices demon- [19] S. M. Bidoki, D. M. Lewis, M. Clark, A. Vakorov, P. A. Miller, strate on : off ratios greater than 10 000 : 1 that hold for over and D. McGorman, “Ink-jet fabrication of electronic components,” 1.2 × 106 s (14 days) and after 4000 flexes. For simplicity, the J. Micromech. Microeng., vol. 17, no. 5, pp. 967–974, May 2007. device’s contacts were evaporated, but there is the potential for [20] J. Wu, K. Mobley, and R. L. McCreery, “Electronic characteristics of fluorene/TiO2 molecular heterojunctions,” J. Chem. Phys., vol. 126, depositing the contacts from a liquid source (such as printing) no. 2, pp. 024 704–024 711, Jan. 2007. for completely roll-to-roll processing. These memristor devices [21] D. B. Strukov, J. L. Borghetti, and R. S. Williams, “Coupled ionic and electronic transport model of thin-film semiconductor memristive behav- have the potential for use as flexible memory components, ior,” Small, vol. 5, no. 9, pp. 1058–1063, May 2009. enabling a paradigm shift in electronics toward inexpensive [22] D. D. Blackstock, W. F. Stickle, C. L. Donley, D. R. Stewart, and lightweight portable physically flexible technologies. R. Stanley Williams, “Internal structure of a molecular junction device: Chemical reduction of PtO2 by Ti evaporation onto an inter- ceding organic monolayer,” J. Phys. Chem. C, Lett., vol. 111, no. 1, ACKNOWLEDGMENT pp. 16–20, 2007. [23] F. Verbakel, S. C. J. Meskers, and R. A. J. Janssen, “Reproductive resistive The authors would like to thank L. Richter for his assistance switching in nonvolatile organic memories,” Appl. Phys. Lett., vol. 91, with the spectroscopic ellipsometry measurements. no. 19, pp. 192 103.1–192 103.3, Nov. 2007.

Authorized licensed use limited to: NIST Researchers. Downloaded on June 3, 2009 at 11:25 from IEEE Xplore. Restrictions apply.